Shopify catalog prep for Black Friday: freeze it early

Which product fields does your sale actually depend on?

Fewer than you think. Your filters read metafields, tags, variant options, vendor, and price. Badges read stock levels and the compare-at price. Sorting rules read inventory.

A field nothing reads can stay messy until January. A field your sale reads has to be right before you build the sale, because the sale gets built on top of whatever it says. That is the whole triage.

So the first thing you write is a list. Six fields, maybe eight. Write it before you open a single product, or you lose an afternoon tidying a description nobody filters on.

How long does a catalog edit take to show up?

Up to 48 hours. Shopify's own troubleshooting asks you to wait 24 to 48 hours for automatic reindexing before you treat a filter change as broken, and suggests a trivial edit and save to force one product through sooner.

That number is what turns a freeze date from discipline into arithmetic. An edit made on the 26th is not guaranteed to be live on the 27th. Count back from that, not from the hour your campaign goes out.

Which explains the Thursday night. The edit saved. It just had not been indexed, and doing it twice did not make it faster.

The eight-week checklist covers where the freeze date sits in the plan. The reason for it is narrower: your storefront runs a day or two behind your admin, and the last two days of November are the worst possible time to find that out.

How do you audit four hundred products without opening four hundred?

Export them and count columns instead. Product metafields appear in the product export as their own columns, but only once their definitions exist under Settings, Custom data. Variant metafields never appear there; those live in the variant bulk editor.

Then six counts. Most of them are a sort and a glance.

A count you can read in one row beats an afternoon of spot checks, because a spot check only ever finds the products you happened to open.

Shopify tags are not case sensitive, so Approved and approved are the same tag and never split into two filter rows. Casing sprawl lives in metafield values and option names instead. Tag sprawl is a different animal: past 5,000 unique tags the admin stops suggesting them as you type, so you retype from memory and invent near-duplicates.

Which of your filters disappear for a shopper in another country?

Three of them, and Shopify documents each one plainly. The price filter does not display for currencies other than your store's default. The product tag filter only displays to shoppers browsing in your default language. Vendor filter values are always drawn from the default language, whatever the shopper picked.

Cross-border made up 16% of global orders across Shopify stores over BFCM 2025, per Shopify's press release published 2 December 2025. Not an edge case.

Nothing in your admin flags a filter that is missing for part of your traffic, and it looks correct in your own browser every time you check.

The test costs two minutes. Open your sale collection in a second market and a second language, then count the filters against what you see at home.

What should a catalog freeze forbid, and who does it bind?

Not everything, or nobody will honor it. Prices move. Inventory moves. Images and descriptions can be improved all the way to launch night.

What stops is narrower, and odd enough that it has to be written down: renaming an option, editing a value a filter reads, introducing a value that does not already exist, and bulk tidy-ups of any kind.

It binds everyone with product edit access, and that list is usually longer than the people in the room. Whoever runs the supplier import is on it. So is whoever fixes a typo when a customer emails about one.

Products still have to launch during those weeks, and they can. One built from values that already exist slots into the filters you have. One carrying a brand new value adds a filter row nobody has previewed and restarts the reindex clock.

Sledge builds filter groups from the metafields, tags, variant options, vendor, and price your catalog already carries. Case and spacing variants collapse on their own; genuinely different spellings you merge yourself, and you pick the label shoppers see. Each value carries a product count from the same request as the results, and a value matching nothing drops off the list rather than sitting there as a dead end. That does not excuse the audit. It does mean the mess you find has somewhere to go that is not a spreadsheet and a prayer.

Seven checks, and then you stop editing

1. Write the field list first. Only what your filters, swatches, badges, and sorting rules read. Six or eight lines.
2. Export the sale set, not the catalog. Define your metafields first, or their columns will not be in the file.
3. Sort each column and count twice. Once for blanks, once for values that differ by a space, a colon, or a plural.
4. Fix in bulk, then stop touching it. Give the reindex its 48 hours before you judge whether the fix worked.
5. Open the storefront in a second currency and a second language, and note which filters vanish.
6. Name the freeze date and the four edits it blocks, then send it to everyone who can edit a product.
7. Agree what happens to a product that launches mid-freeze, before one does.